Vowel pronunciation in Swedish dialects analyzed with R u G/L04 - - PowerPoint PPT Presentation

Vowel pronunciation in Swedish dialects analyzed with RuG/L04

Therese Leinonen Workshop on Research Infrastructure for Linguistic Variation Oslo, 17 September 2009

Outline

• Introduction
• Dialectometric research
• Tools in RuG/L04
• Data
• Acoustic analysis
• Examples of analyses with RuG/L04

Introduction

• RuG/L04: free software for dialectometrics and cartography
• www.let.rug.nl/kleiweg/L04
• developed by Peter Kleiweg, University of Groningen
• Unix, Windows
• no graphical user interface, yet

Dialectometric research

• dialectometry = measuring dialects
• aims: finding dialect areas and describing dialect continuua
• dialectometry emphasizes the aggregate analysis and is data-driven
• statistical methods are used for classifying dialects and exploring dialect con-

tinuua

Tools in RuG/L04

• dialectometric tools, distance measures based on transcribed dialect data:

– Levenshtein distance (string edit distance) – Gewichteter Identitätswert

• statistical tools:

– hierarchical clustering – multidimensional scaling – R interface

• cartography:

– web tool for acquiring geographic data with Google Earth: data points and borders of the studied area – tools for displaying dialectometric results

Data

• SweDia (swedia.ling.gu.se): project carried out by

the universities of Lund, Stockholm and Umeå 1998-2001

• 105 sites in Sweden and Swedish-language parts
• f Finland
• 12 speakers from each site: 3 elderly women, 3

elderly men, 3 young women, 3 young men

• vowels elicited with existing mono- or bi-syllabic

words with the target vowel in a coronal conson- ant context

• 19 words of which the vowels cover the standard

Swedish vowel space: dis, disk, dör, dörr, flytta, lass, lat, leta, lett, lott, lus, låt, lär, lös, nät, sot, särk, söt, typ

Acoustic analysis

• Principal component analysis (PCA) of Bark-filtered vowel spectra (Pols et

al., 1973; Jacobi, 2009)

• two components used as acoustic measure of vowel quality, high correlation

with formants

• each vowel measured at nine points within every vowel segment (starting at

25 % and ending at 75 % of the vowel duration)

• the linguistic distance per vowel between any two varieties is calculated as

the Euclidean distance of the acoustic parameters

• Euclidean distance, where i ranges over the nine sampling points:

distance(x, y) =

• 9
• i=1

((PC1xi − PC1yi)2 + (PC2xi − PC2yi)2)

• the distance between varieties is the average distance of the 19 vowels

RuG/L04: mapdiff

• draws

a map

• f

differences between neighbors

• darker lines indicate a larger differ-

ence

Multidimensional scaling

• method for visualizing and exploring similarities/dissimilarities in data
• with given pair-wise distances positions in a low-dimensional space can be

assigned to data points

• 3 dimensions visualized in red, green and blue → maps where the language

varieties form a continuum (Heeringa, 2004)

RuG/L04: maprgb

RuG/L04: maprgb

• lder speakers

younger speakers

• significantly shorter distances between geographic varieties among younger

speakers than between older speakers (t(96) = 8.4, p < 0.001)

RuG/L04: maplink

• for each pair of sites: measure the dis-

tance of older and younger speakers sep- arately

• distance(olderi, olderj)

> distance(youngeri, youngerj) = convergence(blue)

• distance(olderi, olderj)

< distance(youngeri, youngerj) = divergence(red)

• darker lines indicate larger differences

RuG/L04: maplink

convergence divergence

RuG/L04: mapclust

• displays groupings in data by using

colors, patterns, numbers or sym- bols

• groupings based on hierarchical

clustering (RuG/L04) or manually indexed data

5 clusters using Ward’s method

RuG/L04: mapclust

Thanks to:

Peter Kleiweg for making the software available:

http://www.let.rug.nl/kleiweg/L04/

The SweDia project for making the data available The dialectometric research group in Groningen for comments and discussion

YOU

for listening! References:

Bruce, G., Elert, C.-C., Engstrand, O. and Eriksson, A. (1999), Phonetics and phonology of the Swedish dialects: a project presentation and a database demonstrator, Proceedings of the 14th International Congress of Phonetic Sciences (ICPhS 99), San Francisco, pp. 321–324. Heeringa, W. (2004), Measuring Dialect Pronunciation Differences using Levenshtein Distance, PhD thesis, Rijksuniversiteit Groningen. Jacobi, I. (2009), On Variation and Change in Diphthongs and Long Vowels of Spoken Dutch, PhD thesis, Universiteit van Amsterdam. Nerbonne, J. (2009), Data-driven dialectology, Language and Linguistics Compass 3(1), 175–198. Pols, L. C. W., Tromp, H. R. C. and Plomp, R. (1973), Frequency analysis of Dutch vowels from 50 male speakers, Journal of the Acoustical Society of America 53, 1093–1101. Tabachnik, B. G. and Fidell, L. S. (2007), Using Mulitvariate Statistics, 5th edn, Pearson.